learning_rate: 5e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
training_steps: 2420

Training Loss	Epoch	Step	Validation Loss
0.6639	0.0279	200	0.6427
0.6196	0.0559	400	0.6166
0.5943	0.0838	600	0.5994
0.5920	0.1117	800	0.5880
0.5608	0.1397	1000	0.5791
0.5570	0.1676	1200	0.5704
0.5472	0.1955	1400	0.5651
0.5546	0.2235	1600	0.5587
0.5411	0.2514	1800	0.5539
0.5378	0.2793	2000	0.5510
0.5370	0.3073	2200	0.5497
0.5409	0.3352	2400	0.5495
0.5356	0.3380	2420	0.5495

jeanmidev/trackmania-tmnf-qwen3.5-0.8b

trackmania-tmnf-qwen3.5-0.8b